Abstract: In order to detect the interface bonding quality of the adhesively bonded structure in composite plates, this paper proposed an inversion imaging method for interface weak bonding defects based on ultrasonic guided wave detection technology and Y-Net convolutional neural network. In this paper, the phase velocity dispersion curve and wave structure of the ultrasonic guided wave propagating in adhesively bonded structure in composite plates were calculated, from which the optimal excitation frequency and excitation mode suitable for detection were selected. A data set based on finite element simulation was created. The Y-Net was built, trained, verified and generalized ability tested, while the defect guided wave detection data and reconstruction algorithm for probabilistic inspection of defects (RAPID) imaging results were used as input, and the real bonding quality results were used as label data. Structural similarity index measure (SSIM) and peak signal-to-noise ratio (PSNR) were used to evaluate the inversion ability of Y-Net quantitatively. The experimental system was built, and the adhesively bonded structure in composite plates detection experiment was carried out. The results show that the method proposed in this paper can realize the bonding quality detection by means of inversion imaging of weak bonding defects, and the imaging results can accurately and high-quality characterize the position, shape, size and degree of weak bonding of weak bonding defects and other characteristics.

Key words: composite, guided wave, convolutional neural network, weak bonding defect, inversion imaging